We watched 165 AI pricing pages for 13 weeks. Prices went down more than up.
Between and we re-read every tracked vendor's pricing page once a week and diffed the figures. 1 price move survived confirmation. It was a cut. That is the opposite of the direction most coverage of AI tooling assumes, and it is the whole finding.
Every confirmed move, to scale
One bar per tier that changed. Bars share a single axis, so the deepest cut and the biggest rise are directly comparable. Nothing is averaged across billing periods and nothing is converted between currencies.
- LindyPlus · $49.99 → $29.99-40%
1 cut against 0 rises across 5 tools. Median cut -40%, median rise --. The cuts removed $20 a month from the tiers they touched.
We discarded 181 of 187 observations
This is the part of the dataset we would rather you read first. A weekly diff of vendor pricing pages produces a great deal that looks like news and is not, and a report that published all of it would be reporting its own scraper.
- 121
- New tier appeared. A tier we had not recorded before is indistinguishable from our own extraction getting better at reading the same page. Never published as a vendor change.
- 42
- Price moved, not confirmed. The tier is sold on both halves of a monthly/annual switch, or the old figure is still printed beside it, or the value flipped back and forth between checks. Each is what reading the other half of a billing toggle looks like. Ambiguous, so not counted.
- 18
- Limit reworded. The cap did not move. "240,000 credits per year" became "Includes 240K credits /year": same number, new copywriting.
- 6
- Published. A figure that changed between two reads of the same page, where the previous figure is no longer printed anywhere beside that tier. Every one links to the vendor page it was read from.
The moves in full
| Tool | Tier | Was | Now | Change | Caught |
|---|---|---|---|---|---|
| Lindy | Plus | $49.99 | $29.99 | -40% |
"Caught" is the date our read found the new figure, not the date the vendor changed it. We do not know the second one and do not claim it.
21 limit changes, 3 of them real
Usage caps looked like the busiest field in the dataset and turned out to be the quietest. Nearly all of the movement was the same number retypeset by a copywriter. Only these changed what a customer actually gets.
What this does not show
- It is a sample, not the market. 165 tools we track, not every AI tool that exists, and 5 of them moved a price in this window. Most changed nothing.
- List price is not revenue. A published figure says nothing about discounting, annual commitments, or what anyone actually paid.
- We cannot always separate a vendor test from our own read. A page served differently to different visitors looks identical to a price change. Where the old figure was still printed beside the tier, we counted nothing, which is 42 of the moves we saw.
- Six weeks is short. 13 weekly reads establish a direction, not a trend. We publish every subsequent read to the change log, so this page restates itself as the window grows.
Use these numbers
Free to quote, chart, or rebuild with attribution. Every row links to the vendor page it was read from, the methodology describes exactly how a figure is confirmed, and the feed carries every new move as it is found.
dailyaitools, "AI tool prices moved 1 time in 13 weeks, and 1 was a cut", October 26, 2026. https://dailyaitools.ai/pricing-moves/